Comment by zambelli

I'm working on behavioral fine-tuning of small LLMs. Ie, not fine-tuning or distilling knowledge, but operating practices (TODO lists, scratchpad reflex, etc).